Você está na página 1de 55

Applications and current

Applications and current


trends in Image processing
trends in Image processing
Overview
Overview

Image Concepts and Structures

Compression Techniques

Applications in image processing


Image Concepts and
Image Concepts and
Structures
Structures

Digital image

A 2D array of samples (pixels)

Leel of intensity ! num"er of "its per pixel

Classification

#inary images

Computer graphics

$rayscale images

Color images
%hy do &e process images'
%hy do &e process images'

(ACILITAT) *ICT+,) ST-,A$)

)./A.C)0).T A.D ,)ST-,ATI-.

1IS+ALI2ATI-.

S)$0).TATI-.

C-0*,)SSI-.

APPLICATIONS IN IMAGE
APPLICATIONS IN IMAGE
PROCESSING
PROCESSING

ST)$A.-$,A*/3

ST)$A.AL3SIS

DI$ITAL %AT), 0A,4I.$

I0A$) C-0*,)SSI-.

I0A$) S)$0).TATI-.

I0A$) ,)C-$.ITI-.

I0A$) )./A.C)0).T
STEGANOGRAP!
STEGANOGRAP!

/IDI.$ A DATA I.SID) A I0A$)


-riginal essel
image
)m"edded essel image
LS" ALGORITM
LS" ALGORITM

Least significant "it (LS#) su"stitution


5
)asy to understand and implement
5
+sed in many aaila"le stego tools
)m"edded (ile
Coer (ile
# $ $ $ # $ # $ $ # # $ $ # $ # # $ # $ $ $ $ #
# # $ ... ...
# $ $ $ # $ # # $ # # $ $ # $ # # $ # $ $ $ $ $ ... ...
STEGANAL!SIS
STEGANAL!SIS

T)C/.I6+) -( D)T)CTI-. -(
ST)$- I0A$)S
%igital water mar&ing
%igital water mar&ing

T&o methods for copyright protection


5
Cryptography

Transform digital contents into unperceia"le


format

)ncryption7Decryption
5
%atermar8ing

/ide the copyright control information into digital


content "y modification the content itself
$eneric &atermar8 em"edding scheme
$eneric &atermar8 em"edding scheme
$eneric &atermar8 recoery scheme


&atermar8ed image
,ecoered correct image
-riginal image
Watermarking technique
'() %o 'e Need Compression*
'() %o 'e Need Compression*

T- ,)D+C) T/) 0)0-,3

)((ICI).T +S) -( #A.D%IDT/


T&o types of compression

Lossy compression

Loss less compression


Loss) Compression
Loss) Compression
LOSSY COMPRESSION

HIGH COMPRESSION RATIO

REMOVING THE INORMATION ROM THE SIGNAL

INORMATION LOSS

*redictie (0otion Compensation)

(requency -riented (Transform9 Su":"and)

Importance -riented ((iltering9 #it Allocation9 Sampling9


6uanti;ation)

/y"rid (<*)$9 0*)$9 px=>)


Loss less compression
Loss less compression
LOSSLESS COMPRESSION

REMOVING THE RE!"N!ANT INORMATION


ROM THE IMAGE

LOW COMPRESSION RATIO

NO LOSS

/uffman Coding

Arithmetic Coding

Lempel 2i Algorithm

,un:length Coding
T)pes o+ compression
T)pes o+ compression
algorit(ms
algorit(ms

DCT

D%T

S*I/T

1)CT-, 6+A.TI2ATI-.

(,ACTAL C-0*,)SSI-.
"loc& Trans+orm Encoding
"loc& Trans+orm Encoding
2ig:;ag
6uanti;e
,un:length
Code
/uffman
Code
011010001011101...
DCT
139 144 149 153
144 151 153 156
150 155 160 163
159 161 162 160
original image
DCT
1260 -1 -12 -5
-23 -17 -6 -3
-11 -9 -2 2
-7 -2 0 1
AC components
6uanti;e
;ig;ag
79 0 -2 -1 -1 -1 0 0 -1 0 0 0 0 0 0 0
run:length
code
/uffman
code
10011011100011...
DC component
Block
Encoding
,PEG COMPRESSION
,PEG COMPRESSION
(DCT 6uanti;er )ntropy
)ncoder
6:Ta"le
)ncoding
Ta"le
xxxxxxx
xxxxxxx
xxxxxxx
<*)$
)ncoder
"itstream
IDCT Dequanti;er )ntropy
Decoder
6:Ta"le
Decoding
Ta"le
xxxxxxx
xxxxxxx
xxxxxxx
<*)$
Decoder
"itstream
COMPRESSION RATIO -$.#
COMPRESSION RATIO -$.#
#efore compression
After compression
"loc& Trans+orm Encoding
"loc& Trans+orm Encoding
2ig:;ag
6uanti;e
,un:length
Code
/uffman
Code
011010001011101...
DCT
139 144 149 153
144 151 153 156
150 155 160 163
159 161 162 160
original image
DCT
1260 -1 -12 -5
-23 -17 -6 -3
-11 -9 -2 2
-7 -2 0 1
AC components
6uanti;e
;ig;ag
79 0 -2 -1 -1 -1 0 0 -1 0 0 0 0 0 0 0
run:length
code
/uffman
code
10011011100011...
DC component
Block
Encoding
Image recognition
Image recognition
Image recognition
Image recognition
Need +or "io metric s)stem
Need +or "io metric s)stem

/ac8ers in cryptography

#iometric data is unique and cannot "e


transferred

)xtremely difficult to duplicate


Identi+ication vs veri+ication
Identi+ication vs veri+ication

1erification (one to one)


)g pass&ord erification

Identification (-ne to many)


)g face recognition

.eed for high speed processors to perform


computation
"iometric tec(nolog)
"iometric tec(nolog)

(AC) ,)C-$.ITI-.

(I.$), *,I.T ,)C-$.ITI-.

I,IS ,)C-$.ITI-.

,)TI.A SCA.
'or&ing o+ a /iometric s)stem
'or&ing o+ a /iometric s)stem
"LOC0 %IAGRAM 1OR 1ACE
"LOC0 %IAGRAM 1OR 1ACE
RECOGNITION
RECOGNITION
Input face image
(ace detection
(ace feature
extraction
(eature 0atching
Decision ma8er
-utput result
Face
database
Face
recognitio
n
T(eor) "e(ind
T(eor) "e(ind
1acial Recognition
1acial Recognition

)igenface Technology
Local 1eature Anal)sis. 2345$
/loc&s
T(eor) "e(ind
T(eor) "e(ind
1acial Recognition
1acial Recognition
Iris Anatom) and P()siolog)
Iris Anatom) and P()siolog)

-nly internal organ that is


normally isi"le externally?

Isolated and protected

Surgical ris8s of
modifications

Tra"ecular mesh&or8 of
connectie tissue
Iris Scan Process
Iris Scan Process

0a@or Steps

Isolation of iris tissue


image

(eature encoding
using 2D quadrature
$a"or %aelet
demodulation

Statistical Analyses
Iris Image Isolation
Iris Image Isolation

0inimum AB pixels

CBB:C>B more typical

0onochrome cameras
5
1isi"le light
5
Infrared light

ABB 5 DBB nm

Textures reealed

Integrodifferential
-perator

*upillary and outer


"oundary

$aussian smoothing
operator
Ga/or 'avelet %emodulation
Ga/or 'avelet %emodulation

6uadrature 2D $a"or
&aelets
5
+sed to encode iris features
5
-"tain iris code
5
29B>E phase "its computed
(2F= "ytes)
5
*hase information
5
Amplitude information
depends on too many
factors

Contrast9 illumination9
etcG
More Ga/or6
More Ga/or6

(x
B
9y
B
) specify &aelet position

specify &aelet length

specify a modulation &ae ector &hich can


"e interpreted as polar coordinates as spatial
frequency
( ) ( ) ( ) ( ) H I 2 H 7 7 I
B B B B
2 2
B
2 2
B
) 9 (
y y v x x u i y y x x
e e y x
+ +
=

) 9 (
Statistics
Statistics

The key to iis eco!nition


is the "#i$ue o" # test o"
st#tistic#$ in%epen%ence.
&1'

*ass 5 different eyes

(ail 5 same eyes

/amming Distance (/D)


computation
5
0easure of dissimilarity

J-,
5
Detects Disagreement

A.D
5
*rotects against Corruption

A"out CBB9BBB irises7sec


on a KBB 0/; C*+
1inger print veri+ication s)stem
1inger print veri+ication s)stem
1eature e7traction
1eature e7traction
1eature En(ancement
1eature En(ancement
Origina#
Enhance$
Aut%matic Minutiae !etecti%n&
0inutiae are essentially
terminations and "ifurcations
of the ridge lines that constitute
a fingerprint patternL
Matc(ing algorit(m
Matc(ing algorit(m
Input image
Data "ase
Comparision of minutiae detection
recognition
recognition
E'i(ting S)(tem&
BLBCM (A, N CM (,, (depends on
ealuation scheme)
Arti*icia##)
create$
+i%metric(
Attack at
the
!ata,a(e


Attac&s
Attac&s
Attacking
Via In-ut
P%rt
Multi/iometrics S)stems
Multi/iometrics S)stems

mu#ti-#e ,i%metric trait(L

m%re re#ia,#e

anti.(-%%*ing mea(ure(

a$$re(( the -r%,#em %* n%n.uni/er(a#it)


Image segmentation
Image segmentation

Segmentation9 is to distinguish o"@ects from


"ac8ground
IMAGE SEGMENTATION
IMAGE SEGMENTATION
ALGORITMS
ALGORITMS

CA..3 )D$) D)T)CTI-. 0)T/-D

(+223

(+223 0)A.S

4:0)A.S

#*.
"loc& diagram +or
"loc& diagram +or
segmentation
segmentation
Input image
Input image
Preprocessed image
Preprocessed image
Region grown image
Region grown image
Edge detected image
Edge detected image
Segmented /) +u88) logic
Segmented /) +u88) logic
Segmented /) "PN
Segmented /) "PN
Segmentation /) single seed
Segmentation /) single seed
point
point
Minimum 3 seed points
Minimum 3 seed points

Você também pode gostar